What is it all about? Some simple solutions Consider an audioconference where Consider an audioconference where Use a moderator Use a moderator if one person speaks, all can hear if one person speaks, all can hear a speaker must wait for moderator to call on him or her, even if no a speaker must wait for moderator to call on him or her, even if no one else wants to speak one else wants to speak if more than one person speaks at the same time, both voices are if more than one person speaks at the same time, both voices are Multiple Access garbled garbled what if the moderator what if the moderator ʼ s connection breaks? s connection breaks? How should participants coordinate actions so that How should participants coordinate actions so that Distributed solution Distributed solution the number of messages exchanged per second is maximized the number of messages exchanged per second is maximized speak if no one else is speaking speak if no one else is speaking time spent waiting for a chance to speak is minimized time spent waiting for a chance to speak is minimized but if two speakers are waiting for a third to finish, guarantee but if two speakers are waiting for a third to finish, guarantee An Engineering Approach to Computer Networking An Engineering Approach to Computer Networking collision collision This is the This is the multiple access problem multiple access problem Designing good schemes is surprisingly hard! Designing good schemes is surprisingly hard! Outline Contexts for the multiple access problem Contexts Contexts for the problem Contexts for the problem Broadcast Broadcast transmission medium transmission medium Choices and constraints Choices and constraints message from any transmitter is received by all receivers message from any transmitter is received by all receivers Colliding messages are garbled Colliding messages are garbled Performance metrics Performance metrics Goal Goal Base technologies Base technologies maximize message throughput maximize message throughput Centralized schemes Centralized schemes minimize mean waiting time minimize mean waiting time Distributed schemes Distributed schemes Shows up in five main contexts Shows up in five main contexts
Contexts Solving the problem Outline First, choose a First, choose a base technology base technology Contexts for the problem Contexts for the problem to isolate traffic from different stations to isolate traffic from different stations Choices and constraints Choices and constraints can be in time domain or frequency domain can be in time domain or frequency domain Performance metrics Performance metrics Then, choose how to allocate a limited number of transmission Then, choose how to allocate a limited number of transmission Base technologies Base technologies resources to a larger set of contending users resources to a larger set of contending users Centralized schemes Centralized schemes Distributed schemes Distributed schemes Choices Constraints The parameter ʻ a ʼ Centralized vs. distributed design Centralized vs. distributed design Spectrum scarcity Spectrum scarcity The number of packets sent by a source before the farthest The number of packets sent by a source before the farthest station receives the first bit station receives the first bit is there a moderator or not? is there a moderator or not? radio spectrum is hard to come by radio spectrum is hard to come by in a centralized solution one of the stations is a in a centralized solution one of the stations is a master master and the and the only a few frequencies available for long-distance communication only a few frequencies available for long-distance communication others are others are slaves slaves multiple access schemes must be careful not to waste bandwidth multiple access schemes must be careful not to waste bandwidth master->slave = downlink master->slave = downlink Radio link properties Radio link properties slave->master = uplink slave->master = uplink radio links are error prone radio links are error prone in a distributed solution, all stations are peers in a distributed solution, all stations are peers fading fading Circuit-mode vs. packet-mode Circuit-mode vs. packet-mode multipath interference multipath interference do stations send steady streams or bursts of packets? do stations send steady streams or bursts of packets? hidden terminals hidden terminals with streams, doesn with streams, doesn ʼ t make sense to contend for every packet t make sense to contend for every packet transmitter heard only by a subset of receivers transmitter heard only by a subset of receivers allocate resources to streams allocate resources to streams capture capture with packets, makes sense to contend for every packet to avoid with packets, makes sense to contend for every packet to avoid on collision, station with higher power overpowers the other on collision, station with higher power overpowers the other wasting bandwidth wasting bandwidth lower powered station may never get a chance to be heard lower powered station may never get a chance to be heard
Outline Performance metrics Performance metrics Contexts for the problem Contexts for the problem Normalized throughput Normalized throughput Stability Stability Choices and constraints Choices and constraints fraction of link capacity used to carry non-retransmitted packets fraction of link capacity used to carry non-retransmitted packets with heavy load, is all the time spent on resolving contentions? with heavy load, is all the time spent on resolving contentions? example example => unstable => unstable Performance metrics Performance metrics with no collisions, 1000 packets/sec with no collisions, 1000 packets/sec with a stable algorithm, throughput does not decrease with offered with a stable algorithm, throughput does not decrease with offered Base technologies Base technologies load load with a particular scheme and workload, 250 packets/sec with a particular scheme and workload, 250 packets/sec if infinite number of uncontrolled stations share a link, then if infinite number of uncontrolled stations share a link, then Centralized schemes Centralized schemes => goodput = 0.25 => goodput = 0.25 instability is guaranteed instability is guaranteed Mean delay Mean delay Distributed schemes Distributed schemes but if sources reduce load when overload is detected, can achieve but if sources reduce load when overload is detected, can achieve amount of time a station has to wait before it successfully transmits amount of time a station has to wait before it successfully transmits stability stability a packet a packet Fairness Fairness depends on the load and the characteristics of the medium depends on the load and the characteristics of the medium no single definition no single definition ʻ no-starvation no-starvation ʼ : source eventually gets a chance to send : source eventually gets a chance to send max-min fair share: will study later max-min fair share: will study later Outline Base technologies FDMA Contexts for the problem Contexts for the problem Isolates data from different sources Isolates data from different sources Simplest Simplest Choices and constraints Choices and constraints Three basic choices Three basic choices Best suited for analog links Best suited for analog links Performance metrics Performance metrics Frequency division multiple access (FDMA) Frequency division multiple access (FDMA) Each station has its own frequency band, separated by guard Each station has its own frequency band, separated by guard Time division multiple access (TDMA) Time division multiple access (TDMA) bands bands Base technologies Base technologies Code division multiple access (CDMA) Code division multiple access (CDMA) Receivers tune to the right frequency Receivers tune to the right frequency Centralized schemes Centralized schemes Number of frequencies is limited Number of frequencies is limited Distributed schemes Distributed schemes reduce transmitter power; reuse frequencies in non-adjacent cells reduce transmitter power; reuse frequencies in non-adjacent cells example: voice channel = 30 KHz example: voice channel = 30 KHz 833 channels in 25 MHz band 833 channels in 25 MHz band with hexagonal cells, partition into 118 channels each with hexagonal cells, partition into 118 channels each but with N cells in a city, can get 118N calls => win if N > 7 but with N cells in a city, can get 118N calls => win if N > 7
Recommend
More recommend